Forgery detection using Image Processing

Published on 21 Mar 2022. Written by Athulya Menon

A common problem faced by several newspaper agencies and content creators on the web is that of content forging or image forging. With so much new content hitting the web every single day, how do you make sure an image or news is real and not fake? How do you protect your work from being copied? Well, that’s a common problem for creators now, and something we will try to solve through this project. Here’s an Image Processing project that will help you detect forgery.

Read more..

Project Description

The development of computers, the internet and image processing has led to an increase in digital image forgery. Since such images and infographics are important pools of data, forging them can have severe after-effects and repercussions. Over the past few years, several scientists and researchers have been working on ways to prevent such mishaps from happening. While there are several ways to tamper with a digital image, one of the most popular ways to do so is by the copy-move method, which essentially involves the pasting of image sections from other images to the image in question. During these operations, certain image processing methods such as rotation, scaling, blurring, and noise addition are applied to create more realistic duplicates. A novel forgery detection algorithm makes use of adaptive over-segmentation and feature extraction to detect imperfections in the image. This project will integrate both block-based and point-based techniques to detect the forgery.

Concepts Used

• Fundamentals of Image processing
• Boolean operations on images
• Data segmentation
• Scaling of digital images
• Blurring and Noise addition to images
• Adaptive segmentation methods
• Basics of Feature Extraction
• Block-matching techniques
• MATLAB Basics

Software and Hardware Components

• A desktop or laptop running on an OS higher than Windows XP
• MATLAB R 2012 or higher installed
• Pentium IV 2.4 GHz.chip or higher
• At least 40 GB space in the hard disk
• At least 512 Mb RAM

Project Implementation

1. A simple way to detect forgeries would be to run an exhaustive search, which would involve pixel-wise comparison of the image to a cyclic-shifted version of the same image. However, this would take a lot of time, computational effort and may not always bear fruit, especially if the final image has undergone a lot of modifications.
2. Adaptive Over-Segmentation allows us to segment the data set or input image into blocks.
3. Feature extraction then helps us extract the relevant feature points from each block and set them as block features.
4. These block features are then compared and matched with each other to locate labeled features.
5. Such an algorithm when iterated enough times can help us spot the suspected forgery.
6. This method will not give us the exact location of forgery, but rather give us a localized region within which the forgery has occurred.
7. For further refinement and accuracy regarding the forgery, we should employ the Forgery Region Extraction algorithm.
8. This will help us replace all the feature points with small feature-block superpixels.
9. Next, we will merge all neighboring blocks which have a similar background color or pixel value and group them into merged feature blocks.
10. Then, a morphological operation will be performed on the merged regions to detect the forged regions.
11. As the overlapping blocks have been moved by the same amount of distance or shift, the distance between them remains the same. Hence, for this detection technique to be successful, there must be at least enough similar blocks in the image to connect with each other and form several regions of the same shape within the image.

Kit required to develop Forgery detection using Image Processing:

• No kit required

Technologies you will learn by working on Forgery detection using Image Processing:

• Computer vision
• Image Processing

Any Questions?